Message-ID: <tip-cf6d445f68974d0b15a14cf6021be38a91f2b5d8@git.kernel.org>
Date:	Mon, 29 Feb 2016 03:07:37 -0800
From:	tip-bot for Thomas Gleixner <tipbot@...or.com>
To:	linux-tip-commits@...r.kernel.org
Cc:	kan.liang@...el.com, linux-kernel@...r.kernel.org,
	vincent.weaver@...ne.edu, mingo@...nel.org, acme@...hat.com,
	peterz@...radead.org, torvalds@...ux-foundation.org,
	jacob.jun.pan@...ux.intel.com, jolsa@...hat.com,
	eranian@...gle.com, harish.chegondi@...el.com, hpa@...or.com,
	andi.kleen@...el.com, tglx@...utronix.de, bp@...en8.de
Subject: [tip:perf/core] perf/x86/uncore: Track packages, not per CPU data

Commit-ID:  cf6d445f68974d0b15a14cf6021be38a91f2b5d8
Gitweb:     http://git.kernel.org/tip/cf6d445f68974d0b15a14cf6021be38a91f2b5d8
Author:     Thomas Gleixner <tglx@...utronix.de>
AuthorDate: Mon, 22 Feb 2016 22:19:16 +0000
Committer:  Ingo Molnar <mingo@...nel.org>
CommitDate: Mon, 29 Feb 2016 09:35:18 +0100

perf/x86/uncore: Track packages, not per CPU data

Uncore is a per package facility, but the code tries to mimick a per CPU
facility with completely convoluted constructs.

Simplify the whole machinery by tracking per package information. While at it,
avoid the kfree/alloc dance when a CPU goes offline and online again. There is
no point in freeing the box after it was allocated. We just keep proper
refcounting and the first CPU which comes online in a package does the
initialization/activation of the box.

Signed-off-by: Thomas Gleixner <tglx@...utronix.de>
Signed-off-by: Peter Zijlstra (Intel) <peterz@...radead.org>
Cc: Andi Kleen <andi.kleen@...el.com>
Cc: Arnaldo Carvalho de Melo <acme@...hat.com>
Cc: Borislav Petkov <bp@...en8.de>
Cc: Harish Chegondi <harish.chegondi@...el.com>
Cc: Jacob Pan <jacob.jun.pan@...ux.intel.com>
Cc: Jiri Olsa <jolsa@...hat.com>
Cc: Kan Liang <kan.liang@...el.com>
Cc: Linus Torvalds <torvalds@...ux-foundation.org>
Cc: Peter Zijlstra <peterz@...radead.org>
Cc: Stephane Eranian <eranian@...gle.com>
Cc: Vince Weaver <vincent.weaver@...ne.edu>
Cc: linux-kernel@...r.kernel.org
Link: http://lkml.kernel.org/r/20160222221011.622258933@linutronix.de
Signed-off-by: Ingo Molnar <mingo@...nel.org>
---
 arch/x86/events/intel/uncore.c       | 422 +++++++++++++++--------------------
 arch/x86/events/intel/uncore.h       |  18 +-
 arch/x86/events/intel/uncore_snbep.c |  10 +-
 3 files changed, 198 insertions(+), 252 deletions(-)

diff --git a/arch/x86/events/intel/uncore.c b/arch/x86/events/intel/uncore.c
index 220f81c..096fc0d 100644
--- a/arch/x86/events/intel/uncore.c
+++ b/arch/x86/events/intel/uncore.c
@@ -9,9 +9,9 @@ struct pci_driver *uncore_pci_driver;
 /* pci bus to socket mapping */
 DEFINE_RAW_SPINLOCK(pci2phy_map_lock);
 struct list_head pci2phy_map_head = LIST_HEAD_INIT(pci2phy_map_head);
-struct pci_dev *uncore_extra_pci_dev[UNCORE_SOCKET_MAX][UNCORE_EXTRA_PCI_DEV_MAX];
+struct pci_extra_dev *uncore_extra_pci_dev;
+static int max_packages;
 
-static DEFINE_RAW_SPINLOCK(uncore_box_lock);
 /* mask of cpus that collect uncore events */
 static cpumask_t uncore_cpu_mask;
 
@@ -94,27 +94,7 @@ ssize_t uncore_event_show(struct kobject *kobj,
 
 struct intel_uncore_box *uncore_pmu_to_box(struct intel_uncore_pmu *pmu, int cpu)
 {
-	struct intel_uncore_box *box;
-
-	box = *per_cpu_ptr(pmu->box, cpu);
-	if (box)
-		return box;
-
-	raw_spin_lock(&uncore_box_lock);
-	/* Recheck in lock to handle races. */
-	if (*per_cpu_ptr(pmu->box, cpu))
-		goto out;
-	list_for_each_entry(box, &pmu->box_list, list) {
-		if (box->phys_id == topology_physical_package_id(cpu)) {
-			atomic_inc(&box->refcnt);
-			*per_cpu_ptr(pmu->box, cpu) = box;
-			break;
-		}
-	}
-out:
-	raw_spin_unlock(&uncore_box_lock);
-
-	return *per_cpu_ptr(pmu->box, cpu);
+	return pmu->boxes[topology_logical_package_id(cpu)];
 }
 
 u64 uncore_msr_read_counter(struct intel_uncore_box *box, struct perf_event *event)
@@ -315,9 +295,9 @@ static struct intel_uncore_box *uncore_alloc_box(struct intel_uncore_type *type,
 		raw_spin_lock_init(&box->shared_regs[i].lock);
 
 	uncore_pmu_init_hrtimer(box);
-	atomic_set(&box->refcnt, 1);
 	box->cpu = -1;
-	box->phys_id = -1;
+	box->pci_phys_id = -1;
+	box->pkgid = -1;
 
 	/* set default hrtimer timeout */
 	box->hrtimer_duration = UNCORE_PMU_HRTIMER_INTERVAL;
@@ -774,14 +754,24 @@ static void uncore_pmu_unregister(struct intel_uncore_pmu *pmu)
 	pmu->registered = false;
 }
 
+static void uncore_free_boxes(struct intel_uncore_pmu *pmu)
+{
+	int pkg;
+
+	for (pkg = 0; pkg < max_packages; pkg++)
+		kfree(pmu->boxes[pkg]);
+	kfree(pmu->boxes);
+}
+
 static void __init uncore_type_exit(struct intel_uncore_type *type)
 {
+	struct intel_uncore_pmu *pmu = type->pmus;
 	int i;
 
-	if (type->pmus) {
-		for (i = 0; i < type->num_boxes; i++) {
-			uncore_pmu_unregister(&type->pmus[i]);
-			free_percpu(type->pmus[i].box);
+	if (pmu) {
+		for (i = 0; i < type->num_boxes; i++, pmu++) {
+			uncore_pmu_unregister(pmu);
+			uncore_free_boxes(pmu);
 		}
 		kfree(type->pmus);
 		type->pmus = NULL;
@@ -796,37 +786,36 @@ static void __init uncore_types_exit(struct intel_uncore_type **types)
 		uncore_type_exit(*types);
 }
 
-static int __init uncore_type_init(struct intel_uncore_type *type)
+static int __init uncore_type_init(struct intel_uncore_type *type, bool setid)
 {
 	struct intel_uncore_pmu *pmus;
 	struct attribute_group *attr_group;
 	struct attribute **attrs;
+	size_t size;
 	int i, j;
 
 	pmus = kzalloc(sizeof(*pmus) * type->num_boxes, GFP_KERNEL);
 	if (!pmus)
 		return -ENOMEM;
 
-	type->pmus = pmus;
-
-	type->unconstrainted = (struct event_constraint)
-		__EVENT_CONSTRAINT(0, (1ULL << type->num_counters) - 1,
-				0, type->num_counters, 0, 0);
+	size = max_packages * sizeof(struct intel_uncore_box *);
 
 	for (i = 0; i < type->num_boxes; i++) {
-		pmus[i].func_id = -1;
-		pmus[i].pmu_idx = i;
-		pmus[i].type = type;
-		INIT_LIST_HEAD(&pmus[i].box_list);
-		pmus[i].box = alloc_percpu(struct intel_uncore_box *);
-		if (!pmus[i].box)
+		pmus[i].func_id	= setid ? i : -1;
+		pmus[i].pmu_idx	= i;
+		pmus[i].type	= type;
+		pmus[i].boxes	= kzalloc(size, GFP_KERNEL);
+		if (!pmus[i].boxes)
 			return -ENOMEM;
 	}
 
+	type->pmus = pmus;
+	type->unconstrainted = (struct event_constraint)
+		__EVENT_CONSTRAINT(0, (1ULL << type->num_counters) - 1,
+				0, type->num_counters, 0, 0);
+
 	if (type->event_descs) {
-		i = 0;
-		while (type->event_descs[i].attr.attr.name)
-			i++;
+		for (i = 0; type->event_descs[i].attr.attr.name; i++);
 
 		attr_group = kzalloc(sizeof(struct attribute *) * (i + 1) +
 					sizeof(*attr_group), GFP_KERNEL);
@@ -847,12 +836,13 @@ static int __init uncore_type_init(struct intel_uncore_type *type)
 	return 0;
 }
 
-static int __init uncore_types_init(struct intel_uncore_type **types)
+static int __init
+uncore_types_init(struct intel_uncore_type **types, bool setid)
 {
-	int i, ret;
+	int ret;
 
-	for (i = 0; types[i]; i++) {
-		ret = uncore_type_init(types[i]);
+	for (; *types; types++) {
+		ret = uncore_type_init(*types, setid);
 		if (ret)
 			return ret;
 	}
@@ -864,28 +854,28 @@ static int __init uncore_types_init(struct intel_uncore_type **types)
  */
 static int uncore_pci_probe(struct pci_dev *pdev, const struct pci_device_id *id)
 {
+	struct intel_uncore_type *type;
 	struct intel_uncore_pmu *pmu;
 	struct intel_uncore_box *box;
-	struct intel_uncore_type *type;
-	bool first_box = false;
-	int phys_id, ret;
+	int phys_id, pkg, ret;
 
 	phys_id = uncore_pcibus_to_physid(pdev->bus);
-	if (phys_id < 0 || phys_id >= UNCORE_SOCKET_MAX)
+	if (phys_id < 0)
 		return -ENODEV;
 
+	pkg = topology_phys_to_logical_pkg(phys_id);
+	if (WARN_ON_ONCE(pkg < 0))
+		return -EINVAL;
+
 	if (UNCORE_PCI_DEV_TYPE(id->driver_data) == UNCORE_EXTRA_PCI_DEV) {
 		int idx = UNCORE_PCI_DEV_IDX(id->driver_data);
-		uncore_extra_pci_dev[phys_id][idx] = pdev;
+
+		uncore_extra_pci_dev[pkg].dev[idx] = pdev;
 		pci_set_drvdata(pdev, NULL);
 		return 0;
 	}
 
 	type = uncore_pci_uncores[UNCORE_PCI_DEV_TYPE(id->driver_data)];
-	box = uncore_alloc_box(type, NUMA_NO_NODE);
-	if (!box)
-		return -ENOMEM;
-
 	/*
 	 * for performance monitoring unit with multiple boxes,
 	 * each box has a different function id.
@@ -902,32 +892,35 @@ static int uncore_pci_probe(struct pci_dev *pdev, const struct pci_device_id *id
 			pmu++;
 	}
 
+	if (WARN_ON_ONCE(pmu->boxes[pkg] != NULL))
+		return -EINVAL;
+
+	box = uncore_alloc_box(type, NUMA_NO_NODE);
+	if (!box)
+		return -ENOMEM;
+
 	if (pmu->func_id < 0)
 		pmu->func_id = pdev->devfn;
 	else
 		WARN_ON_ONCE(pmu->func_id != pdev->devfn);
 
-	box->phys_id = phys_id;
+	atomic_inc(&box->refcnt);
+	box->pci_phys_id = phys_id;
+	box->pkgid = pkg;
 	box->pci_dev = pdev;
 	box->pmu = pmu;
 	uncore_box_init(box);
 	pci_set_drvdata(pdev, box);
 
-	raw_spin_lock(&uncore_box_lock);
-	if (list_empty(&pmu->box_list))
-		first_box = true;
-	list_add_tail(&box->list, &pmu->box_list);
-	raw_spin_unlock(&uncore_box_lock);
-
-	if (!first_box)
+	pmu->boxes[pkg] = box;
+	if (atomic_inc_return(&pmu->activeboxes) > 1)
 		return 0;
 
+	/* First active box registers the pmu */
 	ret = uncore_pmu_register(pmu);
 	if (ret) {
 		pci_set_drvdata(pdev, NULL);
-		raw_spin_lock(&uncore_box_lock);
-		list_del(&box->list);
-		raw_spin_unlock(&uncore_box_lock);
+		pmu->boxes[pkg] = NULL;
 		uncore_box_exit(box);
 		kfree(box);
 	}
@@ -938,15 +931,16 @@ static void uncore_pci_remove(struct pci_dev *pdev)
 {
 	struct intel_uncore_box *box = pci_get_drvdata(pdev);
 	struct intel_uncore_pmu *pmu;
-	int i, cpu, phys_id;
-	bool last_box = false;
+	int i, phys_id, pkg;
 
 	phys_id = uncore_pcibus_to_physid(pdev->bus);
+	pkg = topology_phys_to_logical_pkg(phys_id);
+
 	box = pci_get_drvdata(pdev);
 	if (!box) {
 		for (i = 0; i < UNCORE_EXTRA_PCI_DEV_MAX; i++) {
-			if (uncore_extra_pci_dev[phys_id][i] == pdev) {
-				uncore_extra_pci_dev[phys_id][i] = NULL;
+			if (uncore_extra_pci_dev[pkg].dev[i] == pdev) {
+				uncore_extra_pci_dev[pkg].dev[i] = NULL;
 				break;
 			}
 		}
@@ -955,34 +949,20 @@ static void uncore_pci_remove(struct pci_dev *pdev)
 	}
 
 	pmu = box->pmu;
-	if (WARN_ON_ONCE(phys_id != box->phys_id))
+	if (WARN_ON_ONCE(phys_id != box->pci_phys_id))
 		return;
 
 	pci_set_drvdata(pdev, NULL);
-
-	raw_spin_lock(&uncore_box_lock);
-	list_del(&box->list);
-	if (list_empty(&pmu->box_list))
-		last_box = true;
-	raw_spin_unlock(&uncore_box_lock);
-
-	for_each_possible_cpu(cpu) {
-		if (*per_cpu_ptr(pmu->box, cpu) == box) {
-			*per_cpu_ptr(pmu->box, cpu) = NULL;
-			atomic_dec(&box->refcnt);
-		}
-	}
-
-	WARN_ON_ONCE(atomic_read(&box->refcnt) != 1);
+	pmu->boxes[pkg] = NULL;
+	if (atomic_dec_return(&pmu->activeboxes) == 0)
+		uncore_pmu_unregister(pmu);
 	uncore_box_exit(box);
 	kfree(box);
-
-	if (last_box)
-		uncore_pmu_unregister(pmu);
 }
 
 static int __init uncore_pci_init(void)
 {
+	size_t size;
 	int ret;
 
 	switch (boot_cpu_data.x86_model) {
@@ -1025,24 +1005,34 @@ static int __init uncore_pci_init(void)
 	if (ret)
 		return ret;
 
-	ret = uncore_types_init(uncore_pci_uncores);
-	if (ret)
+	size = max_packages * sizeof(struct pci_extra_dev);
+	uncore_extra_pci_dev = kzalloc(size, GFP_KERNEL);
+	if (!uncore_extra_pci_dev) {
+		ret = -ENOMEM;
 		goto err;
+	}
+
+	ret = uncore_types_init(uncore_pci_uncores, false);
+	if (ret)
+		goto errtype;
 
 	uncore_pci_driver->probe = uncore_pci_probe;
 	uncore_pci_driver->remove = uncore_pci_remove;
 
 	ret = pci_register_driver(uncore_pci_driver);
 	if (ret)
-		goto err;
+		goto errtype;
 
 	pcidrv_registered = true;
 	return 0;
 
-err:
+errtype:
 	uncore_types_exit(uncore_pci_uncores);
-	uncore_pci_uncores = empty_uncore;
+	kfree(uncore_extra_pci_dev);
+	uncore_extra_pci_dev = NULL;
 	uncore_free_pcibus_map();
+err:
+	uncore_pci_uncores = empty_uncore;
 	return ret;
 }
 
@@ -1052,110 +1042,81 @@ static void __init uncore_pci_exit(void)
 		pcidrv_registered = false;
 		pci_unregister_driver(uncore_pci_driver);
 		uncore_types_exit(uncore_pci_uncores);
+		kfree(uncore_extra_pci_dev);
 		uncore_free_pcibus_map();
 	}
 }
 
-/* CPU hot plug/unplug are serialized by cpu_add_remove_lock mutex */
-static LIST_HEAD(boxes_to_free);
-
-static void uncore_kfree_boxes(void)
-{
-	struct intel_uncore_box *box;
-
-	while (!list_empty(&boxes_to_free)) {
-		box = list_entry(boxes_to_free.next,
-				 struct intel_uncore_box, list);
-		list_del(&box->list);
-		kfree(box);
-	}
-}
-
 static void uncore_cpu_dying(int cpu)
 {
-	struct intel_uncore_type *type;
+	struct intel_uncore_type *type, **types = uncore_msr_uncores;
 	struct intel_uncore_pmu *pmu;
 	struct intel_uncore_box *box;
-	int i, j;
+	int i, pkg;
 
-	for (i = 0; uncore_msr_uncores[i]; i++) {
-		type = uncore_msr_uncores[i];
-		for (j = 0; j < type->num_boxes; j++) {
-			pmu = &type->pmus[j];
-			box = *per_cpu_ptr(pmu->box, cpu);
-			*per_cpu_ptr(pmu->box, cpu) = NULL;
-			if (box && atomic_dec_and_test(&box->refcnt)) {
-				list_add(&box->list, &boxes_to_free);
+	pkg = topology_logical_package_id(cpu);
+	for (; *types; types++) {
+		type = *types;
+		pmu = type->pmus;
+		for (i = 0; i < type->num_boxes; i++, pmu++) {
+			box = pmu->boxes[pkg];
+			if (box && atomic_dec_return(&box->refcnt) == 0)
 				uncore_box_exit(box);
-			}
 		}
 	}
 }
 
-static int uncore_cpu_starting(int cpu)
+static void uncore_cpu_starting(int cpu, bool init)
 {
-	struct intel_uncore_type *type;
+	struct intel_uncore_type *type, **types = uncore_msr_uncores;
 	struct intel_uncore_pmu *pmu;
-	struct intel_uncore_box *box, *exist;
-	int i, j, k, phys_id;
-
-	phys_id = topology_physical_package_id(cpu);
-
-	for (i = 0; uncore_msr_uncores[i]; i++) {
-		type = uncore_msr_uncores[i];
-		for (j = 0; j < type->num_boxes; j++) {
-			pmu = &type->pmus[j];
-			box = *per_cpu_ptr(pmu->box, cpu);
-			/* called by uncore_cpu_init? */
-			if (box && box->phys_id >= 0) {
-				uncore_box_init(box);
-				continue;
-			}
+	struct intel_uncore_box *box;
+	int i, pkg, ncpus = 1;
 
-			for_each_online_cpu(k) {
-				exist = *per_cpu_ptr(pmu->box, k);
-				if (exist && exist->phys_id == phys_id) {
-					atomic_inc(&exist->refcnt);
-					*per_cpu_ptr(pmu->box, cpu) = exist;
-					if (box) {
-						list_add(&box->list,
-							 &boxes_to_free);
-						box = NULL;
-					}
-					break;
-				}
-			}
+	if (init) {
+		/*
+		 * On init we get the number of online cpus in the package
+		 * and set refcount for all of them.
+		 */
+		ncpus = cpumask_weight(topology_core_cpumask(cpu));
+	}
 
-			if (box) {
-				box->phys_id = phys_id;
+	pkg = topology_logical_package_id(cpu);
+	for (; *types; types++) {
+		type = *types;
+		pmu = type->pmus;
+		for (i = 0; i < type->num_boxes; i++, pmu++) {
+			box = pmu->boxes[pkg];
+			if (!box)
+				continue;
+			/* The first cpu on a package activates the box */
+			if (atomic_add_return(ncpus, &box->refcnt) == ncpus)
 				uncore_box_init(box);
-			}
 		}
 	}
-	return 0;
 }
 
-static int uncore_cpu_prepare(int cpu, int phys_id)
+static int uncore_cpu_prepare(int cpu)
 {
-	struct intel_uncore_type *type;
+	struct intel_uncore_type *type, **types = uncore_msr_uncores;
 	struct intel_uncore_pmu *pmu;
 	struct intel_uncore_box *box;
-	int i, j;
-
-	for (i = 0; uncore_msr_uncores[i]; i++) {
-		type = uncore_msr_uncores[i];
-		for (j = 0; j < type->num_boxes; j++) {
-			pmu = &type->pmus[j];
-			if (pmu->func_id < 0)
-				pmu->func_id = j;
+	int i, pkg;
 
+	pkg = topology_logical_package_id(cpu);
+	for (; *types; types++) {
+		type = *types;
+		pmu = type->pmus;
+		for (i = 0; i < type->num_boxes; i++, pmu++) {
+			if (pmu->boxes[pkg])
+				continue;
+			/* First cpu of a package allocates the box */
 			box = uncore_alloc_box(type, cpu_to_node(cpu));
 			if (!box)
 				return -ENOMEM;
-
 			box->pmu = pmu;
-			box->phys_id = phys_id;
-			*per_cpu_ptr(pmu->box, cpu) = box;
+			box->pkgid = pkg;
+			pmu->boxes[pkg] = box;
 		}
 	}
 	return 0;
@@ -1166,13 +1127,11 @@ static void uncore_change_type_ctx(struct intel_uncore_type *type, int old_cpu,
 {
 	struct intel_uncore_pmu *pmu = type->pmus;
 	struct intel_uncore_box *box;
-	int i;
+	int i, pkg;
 
+	pkg = topology_logical_package_id(old_cpu < 0 ? new_cpu : old_cpu);
 	for (i = 0; i < type->num_boxes; i++, pmu++) {
-		if (old_cpu < 0)
-			box = uncore_pmu_to_box(pmu, new_cpu);
-		else
-			box = uncore_pmu_to_box(pmu, old_cpu);
+		box = pmu->boxes[pkg];
 		if (!box)
 			continue;
 
@@ -1202,27 +1161,20 @@ static void uncore_change_context(struct intel_uncore_type **uncores,
 
 static void uncore_event_exit_cpu(int cpu)
 {
-	int i, phys_id, target;
+	int target;
 
-	/* if exiting cpu is used for collecting uncore events */
+	/* Check if exiting cpu is used for collecting uncore events */
 	if (!cpumask_test_and_clear_cpu(cpu, &uncore_cpu_mask))
 		return;
 
-	/* find a new cpu to collect uncore events */
-	phys_id = topology_physical_package_id(cpu);
-	target = -1;
-	for_each_online_cpu(i) {
-		if (i == cpu)
-			continue;
-		if (phys_id == topology_physical_package_id(i)) {
-			target = i;
-			break;
-		}
-	}
+	/* Find a new cpu to collect uncore events */
+	target = cpumask_any_but(topology_core_cpumask(cpu), cpu);
 
-	/* migrate uncore events to the new cpu */
-	if (target >= 0)
+	/* Migrate uncore events to the new target */
+	if (target < nr_cpu_ids)
 		cpumask_set_cpu(target, &uncore_cpu_mask);
+	else
+		target = -1;
 
 	uncore_change_context(uncore_msr_uncores, cpu, target);
 	uncore_change_context(uncore_pci_uncores, cpu, target);
@@ -1230,13 +1182,15 @@ static void uncore_event_exit_cpu(int cpu)
 
 static void uncore_event_init_cpu(int cpu)
 {
-	int i, phys_id;
+	int target;
 
-	phys_id = topology_physical_package_id(cpu);
-	for_each_cpu(i, &uncore_cpu_mask) {
-		if (phys_id == topology_physical_package_id(i))
-			return;
-	}
+	/*
+	 * Check if there is an online cpu in the package
+	 * which collects uncore events already.
+	 */
+	target = cpumask_any_and(&uncore_cpu_mask, topology_core_cpumask(cpu));
+	if (target < nr_cpu_ids)
+		return;
 
 	cpumask_set_cpu(cpu, &uncore_cpu_mask);
 
@@ -1249,38 +1203,25 @@ static int uncore_cpu_notifier(struct notifier_block *self,
 {
 	unsigned int cpu = (long)hcpu;
 
-	/* allocate/free data structure for uncore box */
 	switch (action & ~CPU_TASKS_FROZEN) {
 	case CPU_UP_PREPARE:
-		return notifier_from_errno(uncore_cpu_prepare(cpu, -1));
+		return notifier_from_errno(uncore_cpu_prepare(cpu));
+
 	case CPU_STARTING:
-		uncore_cpu_starting(cpu);
+		uncore_cpu_starting(cpu, false);
+	case CPU_DOWN_FAILED:
+		uncore_event_init_cpu(cpu);
 		break;
+
 	case CPU_UP_CANCELED:
 	case CPU_DYING:
 		uncore_cpu_dying(cpu);
 		break;
-	case CPU_ONLINE:
-	case CPU_DEAD:
-		uncore_kfree_boxes();
-		break;
-	default:
-		break;
-	}
 
-	/* select the cpu that collects uncore events */
-	switch (action & ~CPU_TASKS_FROZEN) {
-	case CPU_DOWN_FAILED:
-	case CPU_STARTING:
-		uncore_event_init_cpu(cpu);
-		break;
 	case CPU_DOWN_PREPARE:
 		uncore_event_exit_cpu(cpu);
 		break;
-	default:
-		break;
 	}
-
 	return NOTIFY_OK;
 }
 
@@ -1362,7 +1303,7 @@ static int __init uncore_cpu_init(void)
 		return 0;
 	}
 
-	ret = uncore_types_init(uncore_msr_uncores);
+	ret = uncore_types_init(uncore_msr_uncores, true);
 	if (ret)
 		goto err;
 
@@ -1378,39 +1319,34 @@ err:
 
 static void __init uncore_cpu_setup(void *dummy)
 {
-	uncore_cpu_starting(smp_processor_id());
+	uncore_cpu_starting(smp_processor_id(), true);
 }
 
+/* Lazy to avoid allocation of a few bytes for the normal case */
+static __initdata DECLARE_BITMAP(packages, MAX_LOCAL_APIC);
+
 static int __init uncore_cpumask_init(void)
 {
-	int cpu, ret = 0;
-
-	cpu_notifier_register_begin();
+	unsigned int cpu;
 
 	for_each_online_cpu(cpu) {
-		int i, phys_id = topology_physical_package_id(cpu);
+		unsigned int pkg = topology_logical_package_id(cpu);
+		int ret;
 
-		for_each_cpu(i, &uncore_cpu_mask) {
-			if (phys_id == topology_physical_package_id(i)) {
-				phys_id = -1;
-				break;
-			}
-		}
-		if (phys_id < 0)
+		if (test_and_set_bit(pkg, packages))
 			continue;
-
-		ret = uncore_cpu_prepare(cpu, phys_id);
+		/*
+		 * The first online cpu of each package takes the refcounts
+		 * for all other online cpus in that package.
+		 */
+		ret = uncore_cpu_prepare(cpu);
 		if (ret)
-			goto out;
+			return ret;
 		uncore_event_init_cpu(cpu);
+		smp_call_function_single(cpu, uncore_cpu_setup, NULL, 1);
 	}
-	on_each_cpu(uncore_cpu_setup, NULL, 1);
-
 	__register_cpu_notifier(&uncore_cpu_nb);
-
-out:
-	cpu_notifier_register_done();
-	return ret;
+	return 0;
 }
 
 static int __init intel_uncore_init(void)
@@ -1423,22 +1359,26 @@ static int __init intel_uncore_init(void)
 	if (cpu_has_hypervisor)
 		return -ENODEV;
 
+	max_packages = topology_max_packages();
+
 	ret = uncore_pci_init();
 	if (ret)
 		return ret;
 	ret = uncore_cpu_init();
 	if (ret)
-		goto errpci;
+		goto err;
+
+	cpu_notifier_register_begin();
 	ret = uncore_cpumask_init();
 	if (ret)
-		goto errcpu;
-
+		goto err;
+	cpu_notifier_register_done();
 	return 0;
 
-errcpu:
+err:
 	uncore_types_exit(uncore_msr_uncores);
-errpci:
 	uncore_pci_exit();
+	cpu_notifier_register_done();
 	return ret;
 }
 device_initcall(intel_uncore_init);
diff --git a/arch/x86/events/intel/uncore.h b/arch/x86/events/intel/uncore.h
index 479c1e8..01c6793 100644
--- a/arch/x86/events/intel/uncore.h
+++ b/arch/x86/events/intel/uncore.h
@@ -19,11 +19,12 @@
 #define UNCORE_EXTRA_PCI_DEV		0xff
 #define UNCORE_EXTRA_PCI_DEV_MAX	3
 
-/* support up to 8 sockets */
-#define UNCORE_SOCKET_MAX		8
-
 #define UNCORE_EVENT_CONSTRAINT(c, n) EVENT_CONSTRAINT(c, n, 0xff)
 
+struct pci_extra_dev {
+	struct pci_dev *dev[UNCORE_EXTRA_PCI_DEV_MAX];
+};
+
 struct intel_uncore_ops;
 struct intel_uncore_pmu;
 struct intel_uncore_box;
@@ -79,9 +80,9 @@ struct intel_uncore_pmu {
 	int				pmu_idx;
 	int				func_id;
 	bool				registered;
+	atomic_t			activeboxes;
 	struct intel_uncore_type	*type;
-	struct intel_uncore_box		** __percpu box;
-	struct list_head		box_list;
+	struct intel_uncore_box		**boxes;
 };
 
 struct intel_uncore_extra_reg {
@@ -91,7 +92,8 @@ struct intel_uncore_extra_reg {
 };
 
 struct intel_uncore_box {
-	int phys_id;
+	int pci_phys_id;
+	int pkgid;
 	int n_active;	/* number of active events */
 	int n_events;
 	int cpu;	/* cpu to collect events */
@@ -316,7 +318,7 @@ static inline void uncore_box_exit(struct intel_uncore_box *box)
 
 static inline bool uncore_box_is_fake(struct intel_uncore_box *box)
 {
-	return (box->phys_id < 0);
+	return (box->pkgid < 0);
 }
 
 static inline struct intel_uncore_pmu *uncore_event_to_pmu(struct perf_event *event)
@@ -345,7 +347,7 @@ extern struct intel_uncore_type **uncore_pci_uncores;
 extern struct pci_driver *uncore_pci_driver;
 extern raw_spinlock_t pci2phy_map_lock;
 extern struct list_head pci2phy_map_head;
-extern struct pci_dev *uncore_extra_pci_dev[UNCORE_SOCKET_MAX][UNCORE_EXTRA_PCI_DEV_MAX];
+extern struct pci_extra_dev *uncore_extra_pci_dev;
 extern struct event_constraint uncore_constraint_empty;
 
 /* perf_event_intel_uncore_snb.c */
diff --git a/arch/x86/events/intel/uncore_snbep.c b/arch/x86/events/intel/uncore_snbep.c
index d967fcc..93f6bd9 100644
--- a/arch/x86/events/intel/uncore_snbep.c
+++ b/arch/x86/events/intel/uncore_snbep.c
@@ -986,7 +986,9 @@ static void snbep_qpi_enable_event(struct intel_uncore_box *box, struct perf_eve
 
 	if (reg1->idx != EXTRA_REG_NONE) {
 		int idx = box->pmu->pmu_idx + SNBEP_PCI_QPI_PORT0_FILTER;
-		struct pci_dev *filter_pdev = uncore_extra_pci_dev[box->phys_id][idx];
+		int pkg = topology_phys_to_logical_pkg(box->pci_phys_id);
+		struct pci_dev *filter_pdev = uncore_extra_pci_dev[pkg].dev[idx];
+
 		if (filter_pdev) {
 			pci_write_config_dword(filter_pdev, reg1->reg,
 						(u32)reg1->config);
@@ -2520,14 +2522,16 @@ static struct intel_uncore_type *hswep_msr_uncores[] = {
 
 void hswep_uncore_cpu_init(void)
 {
+	int pkg = topology_phys_to_logical_pkg(0);
+
 	if (hswep_uncore_cbox.num_boxes > boot_cpu_data.x86_max_cores)
 		hswep_uncore_cbox.num_boxes = boot_cpu_data.x86_max_cores;
 
 	/* Detect 6-8 core systems with only two SBOXes */
-	if (uncore_extra_pci_dev[0][HSWEP_PCI_PCU_3]) {
+	if (uncore_extra_pci_dev[pkg].dev[HSWEP_PCI_PCU_3]) {
 		u32 capid4;
 
-		pci_read_config_dword(uncore_extra_pci_dev[0][HSWEP_PCI_PCU_3],
+		pci_read_config_dword(uncore_extra_pci_dev[pkg].dev[HSWEP_PCI_PCU_3],
 				      0x94, &capid4);
 		if (((capid4 >> 6) & 0x3) == 0)
 			hswep_uncore_sbox.num_boxes = 2;

Powered by blists - more mailing lists